statsd-opentsdb-request

Pluggable backend for statsd to send data to OpenTSDB via HTTP requests. This backend expects the metrics to be submitted in a "dotted" format (e.g. environment.hostname.metric_name).

Installation

If you do not have Statsd set up, follow this link

Local node_modules installation
npm install statsd-opentsdb-request

If you perhaps prefer to install this globally then use
npm install -g statsd-opentsdb-request

And then update your backends in the Statsd config file by adding [statsd-opentsdb-request] (more details here)

Configuration

Inside your statsd config file (like this one) include following (the only required field is opentsdbHost)

Note that the tags field of a metric sent to OpenTSDB is always autopopulated by hostname tag which contains the host who sent the metric.

opentsdbHost: Address of your OpenTSDB server ,
opentsdbPort: Port of your OpenTSDB server ,
opentsdbPrefix: Prefix for all of the metrics sent to OpenTSDB ,
opentsdb: {
  prefixCounter : Prefix for all counter metrics ,
  prefixTimer : Prefix for all timer metrics ,
  prefixGauge : Prefix for all gauge metrics ,
  prefixSet : Prefix for all set metrics ,
  tags : Key/value store determining tags to be parsed, explained in detail below
}

Example configuration

opentsdbHost: 'localhost/opentsdb' ,
opentsdbPort: '4242' ,
opentsdbPrefix: 'localtest' ,
opentsdb: {
  prefixCounter : 'counters_test' ,
  prefixTimer : 'timers_test' ,
  prefixGauge : 'gauges_test' ,
  prefixSet : 'sets_test' ,
  tags : {
    ".*test" : "test_type",
    "AWS|localhost" : "host_name"
  }
}

tags in detail

To support submitting to OpenTSDB with tags customisable by the user, we can specify them in the mentioned tags as an array of key/value or to be more precise in this case, regex/tag. The key is a regex matching the submitted name of the metric and the value would be the tag name to which it should be assigned. If this did not make a lot of sense please take a look at some examples below.

Regex tags

There is a metric with name statsd.frontendMachine.invalidArgumentException.exception.

You want to aggregate all exceptions which should be sent to OpenTSDB, and you want the tags to be exception_type and machine which in this case are invalidArgumentException and frontendMachine and which would leave you with a metric name of statsd.exception.

The tags config part would look something like this

tags : {
  ".*Machine" : "machine",
  ".*Exception" : "exception_type"
}

Generated Json object which will be sent to OpenTSDB server would look something like

name: statsd.exception,
timestamp : 1501513598,
value: 1,
tags: { machine : "frontendMachine", exception_type : "invalidArgumentException"}

Regex tags with metric index

There are cases when you don't really have a control of a metric name (like it's sent from a 3rd party or simply it's dynamically generated), but you know that it will always be on a certain index (when split by '.' character). There is an easy solution, just suffix your regex with _i_x. That means that the given regexp can be only matched at the given index of a metric name (remember that we are splitting the metric name on dots and we index from 0).

tags : {
  ".*_i_1" : "endpoint",
  ".*Exception_i_3" : "exception_type"
}

So the tags above would match anything at the index 1 and save it under the tag endpoint. Next, we match anything what matches regex .*Exception and is indexed at 3. For example both of the rules would match a metric name

test.backend.this.InvalidException

because the backend is on index 1 and InvalidException matched the regex .*Exception AND is indexed at position 3.

Just to make it clear, the endpoint rules will match the backend below, but nothing will happen to InvalidException because it is on index 2 and not 3.

test.backend.InvalidException.this